基于小波分解的多尺度PCA周期性攻击检测算法

RESEARCH ON INDUSTRIAL CONTROL PERIODIC ATTACK DETEVTION ALGORITHM BASED ON MULTI-SCALE PCA AND WAVELET DECOMPOSITION

摘要:

在工业控制网络中任何异常入侵行为都直接影响现场控制与决策，工控系统的安全检测迫在眉睫。工业控制系统中存在的正弦波攻击，三角波攻击和方波攻击等周期性入侵攻击，这些攻击隐蔽的分布很难被检测出来并且会造成机器的磨损，目前针对该类攻击检测研究较少。针对以上问题，本文首先采用小波分解把数据分解到各个尺度上，然后采用主成分分析进行局部检测，最后把各个尺度的主成分分析组成一个包含各尺度信息的综合主成分分析模型。通过在数据集及自建数据集的周期性攻击实验结果表明，采用该算法比采用单独主成分分析算法进行攻击检测的整体准确率提高7.4%。

Abstract:

There are periodic intrusion attacks in industrial control system, such as sine wave attack, triangle wave attack and square wave attack. It is difficult to detect the hidden distribution of these attacks. In this algorithm, wavelet decomposition is used to decompose the data to each scale, and principal component analysis is used to detect the local data. Then the principal component analysis of each scale is used to form a comprehensive principal component analysis model which contains the information of each scale. Using Mississippi data set and self built data set, the overall detection accuracy of the algorithm is 7.4% higher than that of the single principal component analysis algorithm.